Contrast Enhancement in MRI Using Combined Double Action Contrast Agents and Image Post-Processing in the Breast Cancer Model

Gd- and Fe-based contrast agents reduce T1 and T2 relaxation times, respectively, are frequently used in MRI, providing improved cancer detection. Recently, contrast agents changing both T1/T2 times, based on core/shell nanoparticles, have been introduced. Although advantages of the T1/T2 agents were shown, MR image contrast of cancerous versus normal adjacent tissue induced by these agents has not yet been analyzed in detail as authors considered changes in cancer MR signal or signal-to-noise ratio after contrast injection rather than changes in signal differences between cancer and normal adjacent tissue. Furthermore, the potential advantages of T1/T2 contrast agents using image manipulation such as subtraction or addition have not been yet discussed in detail. Therefore, we performed theoretical calculations of MR signal in a tumor model using T1-weighted, T2-weighted, and combined images for T1-, T2-, and T1/T2-targeted contrast agents. The results from the tumor model are followed by in vivo experiments using core/shell NaDyF4/NaGdF4 nanoparticles as T1/T2 non-targeted contrast agent in the animal model of triple negative breast cancer. The results show that subtraction of T2-weighted from T1-weighted MR images provides additional increase in the tumor contrast: over two-fold in the tumor model and 12% in the in vivo experiment.


Introduction
Magnetic resonance imaging (MRI) provides the best soft tissue contrast among diagnostic imaging modalities such as computed tomography (CT), positron emission tomography (PET), or X-ray. The contrast provided by MRI is based on proton density as well as the T 1 and T 2 relaxation times of an imaged object. MRI technique utilizes these parameters for contrast manipulation by applying proton density, T 1 -weighted (T 1 w) or T 2 -weighted (T 2 w) MR images. MRI contrast may be provided solely by tissues themselves, due to differences in their relaxation times. Contrast agents shortening T 1 and/or T 2 relaxation times improve the detection of small pathologies such as those seen in early stages of cancer. The most commonly used T 2 w MRI pulse sequences are based on spin echo (SE) or gradient echo (GE) with long repetition time (TR) and long echo time (TE) [1]. For fast T 1 w, inversion recovery (IR) pulse can be added before a pulse sequence (e.g., IR-GE) [2].
Tissues with short T 1 relaxation provide positive (hyperintense) contrast, while tissues with short T 2 relaxation provide a lower signal hence negative (hypointense) contrast. Contrast agents, shortening T 1 and/or T 2 , accumulating in tissues of interest, such as tumors, are used to enhance these effects to improve the diagnostic capabilities of MRI, especially for small biological targets [3].
Due to high r 1 relaxivity (r 1 = 1/T 1 ) Gd 3+ -based particles are commonly used as T 1 contrast agents [4] while superparamagnetic iron oxide nanoparticles (SPION) are used as T 2 contrast agents due to their high r 2 relaxivity (r 2 = 1/T 2 ) [5]. The large magnetic moments of T 2 contrast agents cause high signal change per unit of particles, therefore small quantities of SPION are sufficient to obtain strong contrast imaging [6][7][8]. Furthermore, SPIONS are less toxic than Gd 3+ . Their surface can be functionalized enhancing their specificity, biocompatibility, and biodegradability [6, 9,10]. SPION can be employed for targeted drug delivery to cancer cells and other disease sites by synthesizing them with target-specific proteins [6,7,10]. While SPION can be relatively easily functionalized, they reduce mostly T 2 relaxations, providing negative contrast which limits their diagnostic capability [5].
Recently, core/shell nanoparticles have been produced and used as contrast agents due to their capability of reducing both T 1 and T 2 relaxation times. The nanoparticles that reduce both T 1 and T 2 relaxation times have been used as contrast agents to improve diagnosis of various diseases, for example: atherosclerotic plaque in the carotid arteries [11], liver metastases [12], identification of white matter lesions in multiple sclerosis (MS) [13], prostate cancer in animal models [14], or in functional MRI to improve the contrast between brain tissues [15]. The core of the dual contrast agents consisting of inorganic material can be covered with either inorganic materials (e.g., silica or gold) or organic materials (phospholipids, fatty acids, polysaccharides, peptides, or other surfactants and polymers) [3,16]. The advantages of some of these nanoparticles include lower toxicity, greater dispersibility, biocompatibility, easier conjugation with other bioactive molecules, and increased thermal and chemical stability [3,17]. These properties can be controlled by the selection of their composition and size. For example, iron oxide was doped with Gd 3+ , and NaDyF 4 nanoparticles (NPs) were covered with NaGdF 4 showing high r 1 and r 2 relaxivity [18]. In the NaDyF 4 /NaGdF 4 contrast agent, the dysprosium (Dy 3+ ) ion provides T 2 contrast while Gd 3+ provides mostly T 1 contrast [19]. Hybrids of nanocubes with both iron and gadolinium as T 1 /T 2 dual contrast agents were also produced and used in in vivo MRI [20]. It was also shown that the relaxivities of nanoparticles depend on their size [21].
While manipulation of the contrast agent's composition and its size enables higher relaxivity, the maximum concentration hence toxicity remains to be the major obstacle in improving MRI diagnostic capabilities. To deal with this issue, Girard et al. proposed to apply a combination of MRI pulse sequences to utilize changes in both T 1 and T 2 relaxation times due to the presence of dual action (T 1 and T 2 ) contrast agents [22]. The authors used a spin echo (SE) and ultrashort echo (UTE) pulse sequences and calculated optimal pulse sequence parameters (echo time, repetition time, flip angle) for both sequences to obtain improved contrast [22].
So far, reports on contrast agents analyzed solely the changes in tumor signal after contrast agent injection, neglecting changes in the surrounding tissue [23]. However, from the diagnostic perspective, this approach is not optimal, as the tumor signal changes should be compared to signal changes in the adjacent tissues. Recently some authors proposed image post-processing to enhance the signal-to-noise ratio (SNR) of a tumor, defining SNR as the average tumor signal divided by the standard deviation of the noise [23]. While this approach is advantageous, it does not consider signal changes in both tumor and normal adjacent tissue, as contrast agents accumulate in both tumor and normal adjacent tissue. Therefore, we studied and compared changes in signals of both the tumor and normal adjacent tissue rather than the tumor SNR alone. We also performed a simple image manipulation of a phantom tumor model subtracting and adding T 1 w, and T 2 w images when T 1 /T 2 "perfect" (i.e., accumulating in tumor only) targeted contrast agents were used. The results show that subtraction of T 1 w and T 2 w pre-and post-contrast MR images improves the tumor contrast. Theoretical calculations are followed by an example of in vivo MRI in a mouse model of triple negative breast cancer (TNB) at 9.4T using core/shell NaDyF 4 /NaGdF 4 (21.2 nm/0.6 nm) contrast agent.

Theoretical Analysis of a Phantom Tumor Model
For theoretical analysis of tumor contrast enhancement in MRI accomplished with contrast agents, we considered the signal changes in T 1 w and T 2 w MR images caused by the application of T 1 , T 2 , and T 1 /T 2 contrast agents. We assumed that the tumor is located within the normal adjacent tissue and that both the tumor and normal adjacent tissue have an elliptical shape and are homogenous ( Figure 1). We defined tumor contrast (TC) as a ratio of signal from the tumor divided by the signal from adjacent tissue: TC = S tu /S ti , where S tu is the average signal from the tumor, and S ti is the average signal coming from the adjacent tissue. of in vivo MRI in a mouse model of triple negative breast can core/shell NaDyF4/NaGdF4 (21.2 nm/0.6 nm) contrast agent.

Theoretical Analysis of a Phantom Tumor Model
For theoretical analysis of tumor contrast enhancement in contrast agents, we considered the signal changes in T1w and T2 the application of T1, T2, and T1/T2 contrast agents. We assumed within the normal adjacent tissue and that both the tumor and nor an elliptical shape and are homogenous ( Figure 1). We defined ratio of signal from the tumor divided by the signal from adjacent Stu is the average signal from the tumor, and Sti is the average sig jacent tissue. Based on the previous reports on cancer MRI and, in part cancer studies (e.g., [24]) showing about 50% TC enhancement c injection, we considered a "perfect" tumor phantom model. This without influencing factors occurring in vivo (e.g., variable con metabolism, tumor stage, contrast agent concentration etc.,) and M and its parameters. We first assumed that tumor and surroundin tical in both T1w (equal to 100 arbitrary units (au)) and in T2w im also assumed "perfect" targeted contrast agents, accumulated in ther assumed, for T1 contrast agent, the signal increases by 50% not change in the normal adjacent tissue. Similarly, for T2 contra mor decreases by 50% and is unchanged in normal adjacent issu T1/T2 contrast agent, we assumed 50% signal increase and 50% re images, respectively [25,26]. In addition to providing signal value T2w images, we also added and subtracted images (Table 1).  Based on the previous reports on cancer MRI and, in particular, on MRI of breast cancer studies (e.g., [24]) showing about 50% TC enhancement caused by contrast agent injection, we considered a "perfect" tumor phantom model. This allowed us to study TC without influencing factors occurring in vivo (e.g., variable contrast absorption, tumor metabolism, tumor stage, contrast agent concentration etc.,) and MRI pulse sequence type and its parameters. We first assumed that tumor and surrounding tissue signals are identical in both T 1 w (equal to 100 arbitrary units (au)) and in T 2 w images (equal to 70 au). We also assumed "perfect" targeted contrast agents, accumulated in the tumor only. We further assumed, for T 1 contrast agent, the signal increases by 50% in tumor only and does not change in the normal adjacent tissue. Similarly, for T 2 contrast agent: signal from tumor decreases by 50% and is unchanged in normal adjacent issue in T 2 w MR images. For T 1 /T 2 contrast agent, we assumed 50% signal increase and 50% reduction in T 1 w and T 2 w images, respectively [25,26]. In addition to providing signal values obtained from T 1 w and T 2 w images, we also added and subtracted images (Table 1). Table 1. The signal values from the phantom model of tumor (S tu ), normal adjacent tissue (S ti ), and the corresponding contrast, were obtained before and after the application of the "perfect" T 1 , T 2 , and T 1 /T 2 contrast agents. Contrast values (TC) obtained by subtraction (T 1 w − T 2 w) and addition (T 1 w + T 2 w) of signals are also shown.

Animal Model of Breast Cancer
For the mouse model of TNB, we used MDA-MB-231 cells [14]. Cells were cultured in RPMI supplemented with 5% fetal bovine serum (RPMI/5%) and maintained in a humidified 5% CO 2 atmosphere at 37 • C. Cells were harvested by trypsinization (EDTA/trypsin), resuspended in RPMI/5%, and counted (BioRad TC10 counter, Hercules, CA, USA). About 2 × 10 6 cells/mL of RPMI/5% were prepared and kept at 37 • C in a water bath up to the orthotopic implantation. At that point, cells were centrifuged (1500 rpm, 5 min) and resuspended in 20 µL sterile saline, and injected into the left #4 inguinal mammary gland of a 6-week old nu/nu female mouse. Implanted in the mouse mammary gland, MDA-MB-231 cells grew as solid spheroidal tumors over 3 weeks. Once the tumor had reached a size of 50 mm 3 , the mouse (weight 35.6 g, heart rate 310-840 beats/min, body temp 36.5-38.0 • ) was injected via the tail vein with the contrast agent.

Synthesis and Characterisation of the NaDyF 4 /NaGdF 4 Nanoparticles
The details of the NaDyF 4 /NaGdF 4 synthesis are described elsewhere [14]. Briefly, prior to the synthesis of the core/shell NPs, cubic-phase (α) NaGdF 4 NPs were synthesized, which acted as the sacrificial NPs for shell formation [27]. The multi-step process of the cubic (α) phase of sacrificial NaGdF 4 synthesis started from gadolinium oxide, via gadolinium trifluoroacetate, mixed with various chemicals, heating, washing, and cooling cycles. Finally, the NP precipitated and washed with ethanol and dispersed in hexane. To synthesize NaDyF 4 , dysprosium (III) chloride hexahydrate was stirred in oleic acid and 1-octadecene under vacuum for 45 min at 120 • C. Further steps included heating, cooling, and stirring solutions of methanol containing sodium hydroxide and ammonium fluoride. After removing methanol by heating, the previously synthesized sacrificial (α) NaGdF 4 nanoparticles dispersed in 1-octadecene (1 mL) were injected into the solution and stirred for 15 min to form a core-shell nanostructure. The nanoparticles were precipitated and washed with ethanol, and then finally dispersed in hexanes [14].

In Vivo MRI Experiments
A 9.4 T, 21 cm magnet bore MRI system (Bruker, Ettlingen, Germany) was used for the in vivo experiments. A volume radio frequency coil was placed over the animal covering the region of interest (ROI). The animal was imaged before and 15 min after the injection of NaDyF 4 /NaGdF 4 (~21.2 nm/0.6 nm) non-targeted contrast agent. A total of 0.250 mL of the contrast agent was administered via the tail vein (vehicle, 0.9% saline) [14]. For the MRI, anesthesia was induced with 4% isoflurane and maintained with 2.7% isoflurane in 69% N 2 O and 30% O 2 using a vaporizer. The animal experiment was approved by the local Animal Care Committee (number AC13-0202).

MR Imaging Parameters
To collect T 1 w and T 2 w in vivo MRI, we used rapid acquisition with relaxation enhancement (RARE) and multi-echo spin-echo (MESE) pulse sequences, respectively. The sequences were triggered with respiration to reduce possible motion artifacts. Parameters used for the T 1 w RARE pulse sequence were: echo time (TE) 7 ms, repetition time (TR) 750 ms, field of view (FOV) 2.56 × 2.56 cm, slice thickness 1 mm, and matrix 128 × 128. Parameters of the T 2 weighted SE pulse sequence were: TE = 35 ms, TR = 5000 ms, FOV = 2.56 × 2.56 cm, slice thickness 1 mm, and matrix size 128 × 128. The images were collected before and 15 min post injection of the NaDyF 4 /NaGdF 4 (average diameter of 21.2 nm and an average shell thickness of 0.6 nm) core/shell non-targeted contrast agent [14].

Selection of ROIs
The ROIs for the tumor and adjacent tissue were selected manually and an example of this is shown in Figure 2. The green line delineates the adjacent tissue and the red line delineates the tumor. Signal values S tu and S ti were calculated as the average of the signal intensities within ROI. Tumor contrast (TC) was analyzed in T 1 w and T 2 w MR images, as well as in subtracted and added T 1 w and T 2 w images before and after contrast agent injection. The subtraction and addition of images were performed using ImageJ (Bethesda, MD, USA) version 1.53 software in 16-bit format. The images were then converted to an 8-bit image jpeg format (0-255) and the analysis was completed with Python (Centrum Wiskunde & Informatica, Amsterdam, The Netherlands) version 3.6. aterials 2023, 16, x FOR PEER REVIEW For the MRI, anesthesia was induced with 4% isoflurane and maintained wi rane in 69% N2O and 30% O2 using a vaporizer. The animal experiment wa the local Animal Care Committee (number AC13-0202).

MR Imaging Parameters
To collect T1w and T2w in vivo MRI, we used rapid acquisition with hancement (RARE) and multi-echo spin-echo (MESE) pulse sequences, res sequences were triggered with respiration to reduce possible motion artifac used for the T1w RARE pulse sequence were: echo time (TE) 7 ms, repetition ms, field of view (FOV) 2.56 × 2.56 cm, slice thickness 1 mm, and matrix 128 eters of the T2 weighted SE pulse sequence were: TE = 35 ms, TR = 5000 ms 2.56 cm, slice thickness 1 mm, and matrix size 128 × 128. The images were co and 15 min post injection of the NaDyF4/NaGdF4 (average diameter of 21 average shell thickness of 0.6 nm) core/shell non-targeted contrast agent [1

Selection of ROIs
The ROIs for the tumor and adjacent tissue were selected manually an of this is shown in Figure 2. The green line delineates the adjacent tissue an delineates the tumor. Signal values Stu and Sti were calculated as the averag intensities within ROI. Tumor contrast (TC) was analyzed in T1w and T2w well as in subtracted and added T1w and T2w images before and after con jection. The subtraction and addition of images were performed using Ima MD, USA) version 1.53 software in 16-bit format. The images were then con bit image jpeg format (0-255) and the analysis was completed with Pyt Wiskunde & Informatica, Amsterdam, The Netherlands) version 3.6.

Calculations of Contrast in a Tumor Phantom Model
The signals from the tumor and adjacent tissue as well as TC values before and after application of contrast agents are shown in Table 1 T1w − T2w, and T1w + T2w images, respectively. Figure 5A-D show the corresponding images. The highest TC = 2.17 value is visible in the subtracted images ( Figure 5C). TC values change after injection of the T1/T2 contrast agent and are 1.50, 0.50, 3.83, and 1.09 in T1w, T2w, T1w − T2w, and T1w + T2w images, respectively. Figure 6A-D show the corresponding images. The highest TC = 3.83 value is visible in the subtracted images ( Figure 6C).   T1w − T2w, and T1w + T2w images, respectively. Figure 5A-D show the corresponding images. The highest TC = 2.17 value is visible in the subtracted images ( Figure 5C). TC values change after injection of the T1/T2 contrast agent and are 1.50, 0.50, 3.83, and 1.09 in T1w, T2w, T1w − T2w, and T1w + T2w images, respectively. Figure 6A-D show the corresponding images. The highest TC = 3.83 value is visible in the subtracted images ( Figure 6C).     Figures 7 and 8 show the MR images of a mouse bearing TNB tumor pre-injection ( Figure 7) and post-injection (Figure 8) of the NaDyF4/NaGdF4 non-targeted contrast agent. Images marked with a, b, c, and d correspond to T1w, T2w, subtracted T1w − T2w, and added T1w + T2w, respectively. The signal and TC values in each case are shown in Table 2.   Figures 7 and 8 show the MR images of a mouse bearing TNB tumor pre-injection ( Figure 7) and post-injection (Figure 8) of the NaDyF4/NaGdF4 non-targeted contrast agent. Images marked with a, b, c, and d correspond to T1w, T2w, subtracted T1w − T2w, and added T1w + T2w, respectively. The signal and TC values in each case are shown in Table 2. As seen in Table 1 and Figure 3, TC is the same and equals 1.00 before the contrast agent injection in T 1 w and T 2 w images as well as the subtracted and added images.

In Vivo Experiments
The TC values change in T 1 w, T 2 w, T 1 w − T 2 w, and T 1 w + T 2 w after injection of all contrast agents. For the T 1 contrast agent, TC values are 1.50, 1.00, 2.67, and 0.71 in T 1 w, T 2 w, T 1 w − T 2 w, and T 1 w + T 2 w images, respectively. Figure 4A-D show the corresponding images. The highest TC = 2.67 is visible in the subtracted images ( Figure 4C). The TC values change after injection of the T 2 contrast agent and are 1.00, 0.47, 2.17, and 1.07 in T 1 w, T 2 w, T 1 w − T 2 w, and T 1 w + T 2 w images, respectively. Figure 5A-D show the corresponding images. The highest TC = 2.17 value is visible in the subtracted images ( Figure 5C). TC values change after injection of the T 1 /T 2 contrast agent and are 1.50, 0.50, 3.83, and 1.09 in T 1 w, T 2 w, T 1 w − T 2 w, and T 1 w + T 2 w images, respectively. Figure 6A-D show the corresponding images. The highest TC = 3.83 value is visible in the subtracted images ( Figure 6C). Figures 7 and 8 show the MR images of a mouse bearing TNB tumor pre-injection ( Figure 7) and post-injection (Figure 8) of the NaDyF 4 /NaGdF 4 non-targeted contrast agent. Images marked with a, b, c, and d correspond to T 1 w, T 2 w, subtracted T 1 w − T 2 w, and added T 1 w + T 2 w, respectively. The signal and TC values in each case are shown in Table 2.      The results of in vivo experiments pre-injection (Figure 7) of the contrast agent show the highest TC value for subtracted T 1 w − T 2 w images (TC = 1.52) when compared to T 1 w, T 2 w and T 1 w + T 2 w images with TC equal to 1.15, 0.33, and 1.00, respectively. Injection of the non-targeted NaDyF 4 /NaGdF 4 contrast agent into a mouse bearing TNB cancer increased the tumor contrast in the T 1 w image ( Figure 8A), and decreased in the T 2 w image ( Figure 8B), as expected. Image of the added T 1 w + T 2 w ( Figure 8D) showed lower tumor contrast (1.09) compared to the T 1 w image (1.24) and higher contrast compared to that in the T 2 w image (0.32). The subtracted T 1 w − T 2 w image ( Figure 8C) showed that the TC was the highest (1.56).

Discussion
Increasing tumor contrast is an ever-evolving pursuit for diagnostic imaging, specifically in MRI. Increasing the tumor contrast can lead to earlier tumor detection. While MRI techniques such as T 1 w and T 2 w imaging enable visualization of tumor regions, they are not sufficient for the detection of small tumors, and tumors with similar signals as surrounding tissues. To overcome this limitation T 1 , T 2 , and T 1 /T 2 contrast agents have been applied. Of particular interest are T 1 /T 2 contrast agents, as their surface can be used for conjugation with various delivery vehicles, such as antibody [28][29][30]. However, T 1 /T 2 contrast agents provide lower TC in T 1 w images compared to T 1 -only contrast agents as their T 1 shortening effect is usually smaller than that of the "pure" T 1 contrast. Similarly, shortening T 2 is smaller for T 1 /T 2 contrast agents than for "perfect" T 2 contrasts, which reduce T 2 only. These effects diminish the potential capabilities of the T 1 /T 2 contrast agents.
The result for the phantom model showed that the application of the T 1 /T 2 contrast agent alone improved tumor contrast by 50% (from 1.0 to 1.5) on T 1 -weighted images and reduced contrast by 50% (from 1.0 to 0.5) on T 2 w images. The image subtraction additionally improved contrast over two-fold (from 1.50 to 3.83) when compared to T 1 -weigthed. For the in vivo experiments, the corresponding values were: increase by 8% (from 1.15 to 1.24) in T 1 -weighted images, decrease by 1% (from 0.33 to 0.32) in T 2 -weighted images, and additional improvement due to image subtraction by 12% (from 1.24 to 1.56). Overall, the results showed that the image contrast enhancement caused by the T 1 /T 2 contrast agents can be additionally improved by the application of image subtraction. Furthermore, TC is the highest for the T 1 /T 2 contrast agent. Subsequently, potential drawbacks associated with lower contrast provided by the T 1 /T 2 contrast agents can be rectified.
The differences between the phantom tumor model and in vivo TC results are due to the fact that our phantom tumor model assumes the application of the "perfect" targeted contrast agents, and our in vivo experiments use a non-targeted contrast agent. Application of the non-targeted contrast agent causes similar changes in signal in both tumor and normal adjacent tissue, while the phantom model shows signal changes in tumor only.
In the presented studies only one type of NPs was used to show the general concept of image analysis. The NaDyF 4 /NaGdF 4 NPs used in the in vivo study had core/shell size of 21.2 nm/0.6 nm. However, we expect the observed increase in TC would be achieved independently of the size of the NP. NPs with higher r1 and r2 would provide even more evident increase in TC. The increase in r 1 and r 2 can be achieved by increasing both the shell and core sizes. In the case of the NaDyF 4 /NaGdF 4 , increase in the Gd shell size would increase r1, hence providing desired improvement in TC. The independence of the TC enhancement due to the applied image post-processing is also supported by the theoretical analysis of the phantom model, where an arbitrary contrast was considered. However, further studies with other NPs are required to find their optimum parameters (e.g., core/shell ratio and size) to obtain maximum TC.

Conclusions
The results show that the T 1 /T 2 contrast agents can indeed provide improved diagnostic capability over the T 1 and T 2 contrast agents alone. However, to fully benefit from its application, the contrast should provide as strong as possible T 1 and T 2 shortening (maximal r 1 and r 2 ), should be targeted, and then T 1 w and T 2 w images should be collected and subtracted. These improvements hold the potential to enhance the diagnostic capabilities of contrast enhanced MRI.